An In-Depth Examination of Java I/O Performance
and
Possible Tuning Strategies
Kai Xu
xuk@cs.wisc.edu
Hongfei Guo guo@cs.wisc.edu
Abstract:
There is a growing interest in using Java for high-performance computing
because of the many advantages that Java offers as a programming
language. To be useful as a language for high-performance computing
however, Java must not only have good support for computation, but must
also be able to provide high-performance file I/O. In this paper, we first
examine possible strategies for doing Java I/O. Then we design and conduct
a series of performance experiments accordingly using C/C++ as a
comparison group. Based on the experimental results and analysis, we reach
our conclusions: Java raw I/O is slower than C/C++, since system calls in
Java are more expensive; buffering improves Java I/O performance, for it
reduces system calls, yet there is no big gain for larger buffer
size; direct buffering is better than the Java-provided buffered I/O
classes, since the user can tailor it for his own needs; increasing the
operation size helps I/O performance without overheads; I/O-related system
calls implemented within Java native methods are cheap, while the overhead
of calling Java native methods is rather high. When the number of JNI
calls is reduced properly, a performance comparable to C/C++ can be
achieved.
The paper is available as:
Postscript
,
MS-WORD
Parts of software used in our experiments:
1. Benchmarks for Java: cs736_bench_java.tar.gz
2. Benchmarks for C/C++: cs736_bench_c.tar.gz
3. Profiler for Java:
PerfAnal
PowerPoint Slides are also available here: cs736_report.ppt